lumen

lumen provides hypothesis tests, confidence intervals, and selected statistical distribution utilities used in the DescToolsX ecosystem. It is designed as a focused statistical companion package: methodologically transparent, API-consistent, and suitable for use in applied statistical workflows.

The package is currently under active development.

Installation

You can install the development version from GitHub:

# install.packages("remotes")
remotes::install_github("AndriSignorell/lumen")

The package requires R >= 4.2.0.

Scope

lumen collects statistical procedures that are commonly useful in exploratory analysis, inference, and methodological comparison. The current package includes:

• hypothesis tests for goodness-of-fit, normality, stationarity, randomness, contingency tables, marginal homogeneity, and nonparametric group comparisons
• confidence intervals for proportions, differences and ratios of proportions, means, medians, variances, correlations, regression coefficients, and related quantities
• post-hoc procedures and multiple-comparison helpers
• selected distribution functions, including extreme-value, Dirichlet, Gompertz, triangular, Benford, and order-statistic distributions
• bootstrap confidence interval helpers

Examples

Binomial confidence intervals

library(lumen)

binomCI(x = 37, n = 43, method = "wilson")
binomCI(x = 42, n = 43, method = "clopper-pearson")

Goodness-of-fit testing

x <- rnorm(50)
andersonDarlingTest(x, null = "pnorm", mean = mean(x), sd = sd(x), estimated = TRUE)

Nonparametric tests

x <- c(1.1, 1.4, 1.6, 2.0, 2.2)
y <- c(1.0, 1.2, 1.3, 1.7, 1.9)

siegelTukeyTest(x, y)

Contingency-table tests

tab <- matrix(c(8, 14, 1, 3), nrow = 2)
barnardTest(tab)

Bootstrap confidence intervals

set.seed(1984)
bootCI(mtcars$mpg, FUN = mean, na.rm = TRUE, bci.method = "basic")

Design principles

lumen follows the broader DescToolsX design philosophy:

• predictable lowerCamelCase function names
• explicit argument validation
• transparent method choices
• clean separation between user-facing interfaces and computational engines
• compatibility with familiar base R idioms where appropriate

Most hypothesis tests return objects compatible with the standard htest interface. Confidence interval functions generally return compact vectors or matrices with estimates and interval bounds.

Selected functions

Hypothesis tests

• andersonDarlingTest() — Anderson-Darling goodness-of-fit test
• bartelsRankTest() — Bartels rank test for randomness
• barnardTest() — Barnard's unconditional test for 2 x 2 tables
• bhapkarTest() — Bhapkar marginal homogeneity test
• breslowDayTest() — Breslow-Day test for homogeneity of odds ratios
• cochranArmitageTest() — Cochran-Armitage trend test
• cochranQTest() — Cochran's Q test
• cramerVonMisesTest() — Cramer-von Mises goodness-of-fit test
• durbinWatsonTest() — Durbin-Watson test
• jarqueBeraTest() — Jarque-Bera normality test
• kpssTest() — KPSS stationarity test
• leveneTest() — Levene test for equality of variances
• lillieTest() — Lilliefors normality test
• siegelTukeyTest() — Siegel-Tukey test for scale differences
• stuartMaxwellTest() — Stuart-Maxwell marginal homogeneity test
• woolfTest() — Woolf test for homogeneity of odds ratios

Confidence intervals

• binomCI() — confidence intervals for binomial proportions
• binomDiffCI() — confidence intervals for differences of binomial proportions
• binomRatioCI() — confidence intervals for ratios of binomial proportions
• bootCI() — bootstrap confidence intervals
• corCI() — confidence intervals for correlations
• meanCI() — confidence intervals for means
• medianCI() — confidence intervals for medians
• multinomCI() — confidence intervals for multinomial proportions
• poissonCI() — confidence intervals for Poisson rates
• quantileCI() — confidence intervals for quantiles
• varCI() — confidence intervals for variances

Distributions and utilities

• dgev(), pgev(), qgev(), rgev() — generalized extreme value distribution
• dgpd(), pgpd(), qgpd(), rgpd() — generalized Pareto distribution
• dgumbel(), pgumbel(), qgumbel(), rgumbel() — Gumbel distribution
• ddirichlet(), pdirichlet(), qdirichlet(), rdirichlet() — Dirichlet distribution
• dtri() — triangular distribution
• dbenford() — Benford distribution
• scores() — score generation helper

Dependencies

lumen imports several packages used for statistical computation and infrastructure, including boot, aurora, bedrock, Exact, mvtnorm, stats, withr, and gld. C++ support is provided through Rcpp, RcppParallel, and RcppArmadillo.

Documentation

The development documentation is available at:

https://andrisignorell.github.io/lumen/

The source repository is available at:

https://github.com/AndriSignorell/lumen/

Issues and feature requests can be submitted at:

https://github.com/AndriSignorell/lumen/issues

License

lumen is released under GPL (>= 2).

Status

This package is experimental and versioned as 0.0.0.907. Interfaces may still change before a stable release.